[MASantos]

So what is the buffer for?? Is it the power supply if you use a mint to power the dac? Is it optional if you are doing a standalone dac?

 

[Francis_Vaughan]

You might like to have a look over at DIYAudio in the digital group. In particular this thread:

http://www.diyaudio.com/forums/showt...threadid=64738

The pointers by Guido Tent are especially valuable, and the articles on board layout and grounding a must read. Board layout has been covered quite a bit of late, so time spent browsing and reading would be very well spent.

There are a few issues with your layout as it currently stands. This is all about the layout for the 2702. The power and especially the ground trace layout is going to lead to all sorts of problems. It is important to realise that the different power and ground pins out of the chip lead very different lives. They serve different circuits in the chip, and each will have its own mess of noise, and sensitivity to noise. For instance the USB ground is on the same thin little trace as the clock generator ground, which then winds its way back to a tie to the groundplane though a single little via. Every bit of current thought the digital part of the device must traverse this path at some point, and you will have crosstalk between the various parts, plus some reasonable chance of coupling into the analog ground, as it wends its serpentine way nearby. The long trace taking AGND to its via is really worrying. You have the opportunity to pick up all sorts of noise here. You need to look to individual vias tying each ground right to the ground plane. You cant allow any loop area.

Also the rather long traces to and from C13 give me worry. There seems to be quite a bit of scope for tightening up the layout to pull the bypass capacitors next to the pins they are serving, rather than sitting out at the end of an inductor, where their utility is seriously compromised.

 

[Alf]

Quote:

Originally Posted by MASantos So what is the buffer for?? Is it the power supply if you use a mint to power the dac? Is it optional if you are doing a standalone dac?

It is for MINT/PIMETA. The DAC uses V- for ground while MINT/PIMETA have virtual ground. Therefore you cannot use the amp's V+ and V- to power the DAC. Also you cannot connect the DAC to V+ and the virtual ground because the TLE2426 can only handle up to 20mA. BUF634 is meant to solve this problem.

If you would like to mate the DAC with your amp, then you should use BUF634 and leave out MAX1722 (and some other parts too).

If you plan to use batteries <5.2V or USB for powering the DAC, then you should use MAX1722 and leave out BUF624.

 

[skyskraper]

Quote:

I am too interested in a high-end DAC design to complement my EMU0404 at home. However I am afraid that this is out of the project's scope.

mmm a nice simple spdif (and torx for those too cheap to buy a spdif source) dac would be very tasty.

 

[Voodoochile]

Guys, please save all the "interested in a board" comments for an actual group buy interest thread, please. Alf does know about the group buy drill, as noted by edwood. So if it comes to that, it will be a seperate thread.

Let's keep this thread for dialog regarding the properties of the project.

 

[Alf]

Quote:

Originally Posted by Francis_Vaughan There are a few issues with your layout as it currently stands. This is all about the layout for the 2702.

Thank you Francis. You are making very valid comments indeed. I will make changes tonight and post a new layout.

By the way, this part of the layout has not changed since Rev A and no one reported the noise issues you mentioned. I wonder whether there will be any audible improvement. I look forward to comparison tests.

 

[Francis_Vaughan]

You won't see this stuff show up as what is heard as a raised noise floor, rather it shows up as our old friend jitter in the digital parts, or actual distortion products. Noise still, where noise is defined as "that which is not part of the desired signal."

Do read this one:

http://www.tentlabs.com/Info/Article...decoupling.pdf

 

[cetoole]

If you are able to get a large quantity of MAX1722 IC's to make available to headfiers, along with the Wima MKS-02 caps, that would be great. For the origional USB DAC, I ended up having to resort to samples for these two parts, say what you will, there just isnt any availability in the States for these parts, and ordering from Farnell isnt really an option. While some here might not like it, if it was possible to add an optional stage using a DC-DC converter, so that instead of the DAC being powered from the Mint/Pimeta, the Mint/Pimeta is powered from regulated USB power, so that people arn't required to use batteries at all, saving vaulable space, and giving a runtime just as long as the laptop, that would be something that I would be very interested in seeing. Providing that the regulation was adequate, it would make the perfect portable laptop headphone amp/dac rig, as it would still fit in the Hammond 1455C802, along with a mint, with no need for an external power source.

 

[MASantos]

Quote:

Originally Posted by cetoole If you are able to get a large quantity of MAX1722 IC's to make available to headfiers, along with the Wima MKS-02 caps, that would be great. For the origional USB DAC, I ended up having to resort to samples for these two parts, say what you will, there just isnt any availability in the States for these parts, and ordering from Farnell isnt really an option. While some here might not like it, if it was possible to add an optional stage using a DC-DC converter, so that instead of the DAC being powered from the Mint/Pimeta, the Mint/Pimeta is powered from regulated USB power, so that people arn't required to use batteries at all, saving vaulable space, and giving a runtime just as long as the laptop, that would be something that I would be very interested in seeing. Providing that the regulation was adequate, it would make the perfect portable laptop headphone amp/dac rig, as it would still fit in the Hammond 1455C802, along with a mint, with no need for an external power source.

The idea of poweriing the amp from the USB power is defenitely interesting, saving both space in the amp and batteries. But does USB have wnough power to feed both the dac and a mint or pimeta?? How would that affect SQ and Thd?

Then the best thing to do is to design 2 separate boards

, one that is a dac powered from a mint, for those who want the best sound quality and another which is powered from the USB cable, saving space in the enclosure and helping with portability!!. Both designs would save space in the PCB by removing components not needed!
the second option would fit in the smallest hamond enclosure( the aluminium ones)

 

[grawk]

I'd also be really interested in a variant on this board that uses one of the DAC chips that can output balanced. It'd be a really nice thing to have stuffed inside a balanced m3 or dynamid...

 

[Botch]

Lots of ideas here but in a way I think we are getting away from the original scope of the project - to have a simple, small USB DAC. Something that Guzzler & the original team did a great job on.

The addition of the interface for MINT is an advantage and only adds one IC but do we really need to complicate the design process by having 2 board designs?

The only improvements that I would like to see are related to the grounding and decreasing the noise floor as per what Francis has posted. This DAC was not designed as a high end DAC, just a easy to build solutiuon and personally I like the simplicity of this design.

Botch.

 

[Alf]

I see a conflict of requirements here. The original revision's design goal was to produce a simple DAC powered from a small battery pack (1 or 2 cells). A lot people asked for a USB powered solution and Guzzler added this option in Rev A.

It seems to me that no one actually wants to hassle with batteries. People use either an external power supply or USB power. When I started thinking about the new revision, I wanted to ditch the low voltage battery section completely. However, I was not sure whether it was a good plan and eventually reverted to the original design.

OK I have to ask you a question now. How many people out there actually using batteries with Rev A? Is there anyone here who strongly advocates this option?

If my guess about batteries is not far off, we have a whole lot of new possibilities. MAX1722 is only required because of the low voltage option. If we ditch this option, then we do not need to use this much hated part.

Here is what I suggest:
* Replace MAX1722 with another DC/DC step-up converter that can boost 5V to around 7-10V.
* Add an additional regulating section that would clean up noise coming from USB and the switch even further.
* Add an option to power an amp from the DAC

Are there any power supply experts here? What can be done to improve the USB option and still fit the board? Can USB be cleaned to an acceptable level?


If this is a go, then I would need your help with selecting the new step up converter and the regulator. Both of them must be:
* (Very) low noise
* Small size but easy to solder
* Available from Farnell and Digikey.

What do you think guys?

 

[Nisbeth]

Allright, if it isn't too late, lets try to nip this "feature-request" discussion in the bud

What I'm hearing people ask for are essentially three different things.

1) Small, portable, bus-powered USB-DAC to feed external amp. Enclosure target: Hammond 1455C801

2) Portable, self-powered USB-DAC/MINT-style amp combo. Enclosure target: Hammond 1455C1201 (or smaller)

3) High-end USB-dac using the I2S-output from a PCM290x to feed an HQ-dac IC. Balanced output preferred. Enclosure target: Probably a Hammond 1455J or larger. (1455K-type for balanced output due to the height of the XLR-connectors).



My evaluation and recommendations are as follows:

1) This is the original scope of this project and the one the thread should continue focusing on. My only comments on what's been achieved so far are that a) even if the board should be designed for the Hammond enclosure, in the interest of versatility on-board connectors should be limited to one edge of the PCB and b) the amp-power section should be retained if the board space allows it (also in the interest of versatility).

2) A very nice concept which i believe many would be looking for, but I still think it deserves a separate thread. Not least because there are a lot of related projects which could be combined, including dougis' LiPo-project, tangent's MINT-revision and the other assorted AD8397-experiments from headwize and of course this project. Development of this design and 1) could easily run parallel to each other but with separate threads and two different persons to coordinate development because they might easily end up with conflicting requirements.

3) A great idea, but completely outsidde the scope of this thread. Start a new thread and use that for feature-requests etc. If necessary, ask Guzzler if he'll donate the files for the USB/SPDIF-board and anything else related he might have lying about to help start the project. There are a few others who have been exploring options like this before and they might be willing to chime in as well

My $0.02 for now

/U.

 

[Alf]

Here is the updated version of the board layout:

Board v2

Changes:
* Moved C18 closer to the IC and changed it to 1206 (DaKi][er)
* Added vias tying each ground to the ground plane (Francis)
* Moved the bypass caps closer to IC pins they are serving (Francis)
* Added an option to replace BUF634 with OPA511 (Nisbeth)

The board is still based on MAX1722. I will replace it as soon as we agree on what we replace it with.

 

[Francis_Vaughan]

I think you still need to push things on the analog side a bit further. Think of the problem like this. You need to connect the power to ground at RF as closely as possible. 0 Volts is easy - just drop a via to the ground plane right on the pin. Power is harder - you need to put a capacitor in the way. But after that the layout is the same problem - get the capacitor right onto the pin, and then the other side straight to the groundplane. You should be able to turn most of the bypassing capacitors around so that they can be lined up along the edge of the chip, with essentially no distance to the pin they serve. The layout seems to have a curious idea that the ground pin for a functional unit should go to the groundplane through the same via as the bypass capacitor for that unit. Thus forcing extra (and bad) trace length. Don't do this. Push all the grounds straight down into the groundplane.

Thus C10,11,13,14 all turn through 90 degees, all right up against the chip, and a via straight down on the other side. The analog ground pins get vias straight down. C12 get as close as possible after.

I notice that there is no seperate bypass capacitor for pins 2 and 4. Egads! Pin 2 is power for the clock generator. Junk on Vdd (pin 4) wil couple straight into the clock - jitter here we come. Get another capacitor right onto pin 2. Same deal as above, via straight down on the other side.
C18 will have to move, but that is no big deal, rotate it 90 degrees, and move the USB coupling resistors. You are going to end up with a transformer here anyway (trust me) so they are not important. There may be a case for supplying each of these power pins via a ferrite bead to keep them seperate.

The crystal is a bit of a mess too. Long traces to the capacitors. Remember that the crystal oscillator determines a large fraction of the jitter in the system. Any signal pickup on those traces will directly cause jitter in the signal - since the signal will modulate the oscillator signal before the comparator from which the internal timing clock is generated. Thus any AC, and worst of all, audio frequency signals, that couple into those traces will cause signal correlated jitter. Lay the ocsillator out with as small a trace length as possible. The tends to mean laying it out like the circuit is typically drawn. Capacitors either side of the crystal with minimum trace lengths. Resistor parallel to the crystal. Right on top of the chip. Get those loop areas down.